<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=GB2312">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Dear All,<br>
I hope you're having a nice day. <br>
Based on ksp ex45, a 3D Poisson solver with non-uniform grid is
coded. The PETSc options I used is:<br>
<i>mpiexec -np 32 ./ex45 -pc_type gamg -ksp_type cg -pc_gamg_type
agg -pc_gamg_agg_nsmooths 1 -mg_levels_ksp_max_it 1
-mg_levels_ksp_type richardson -ksp_rtol 1.0e-7</i><br>
There are some problems:<br>
1, if the mesh is very coarse with very small amount of the grid,
the code runs well;<br>
2, if the mesh is fine, an error message comes up saying
'un-symmetric graph'. It suggests me to use '-pc_gamg_sym_graph
true' or '-pc_gamg_thredhold 0.0'. <br>
a) if '-pc_gamg_sym_graph true' is used, the code runs but blows
up very quickly with crazy norm;<br>
b) if '-pc_gamg_thredhold 0.0' is used, the code stops on the
KSPSolve() forever;<br>
3, because of the 'un-symmetric graph' error, it reminds me that the
matrix may not be symmetric, which indicates that '-ksp_type cg' may
not be a good option. Therefore, I changed it to '-ksp_type gmres'.
It makes the code run with 'moderate fine' mesh. However, it stops
at KSPSolve() also with fine mesh. <br>
a) does there any other '-ksp_type' fit better for this case?<br>
b) does multigrid preconditioner work well for gmres?<br>
<br>
thanks,<br>
Alan<br>
</body>
</html>